The Research Of Face Recognition Method Based On Deep Learning

Face recognition is one of the basic technologies for human-computer interaction.Many research projects and commercial products have put face recognition into use.Deep learning as an emerging field of machine learning has attracted much attention from researchers.Because deep learning has better expression and generalization performance,more and more researchers apply deep learning to face recognition.The current mainstream face recognition algorithm based on deep learning has achieved high recognition rate in standard data sets.However,the model training usually requires a large-scale sample set,and the traditional algorithm is generally used in the face detection stage,and the accuracy of detecting the face with a large deflection angle is low.In this paper,the face detection based on deep learning method is used to improve the detection success rate of the face image with large inclination angle.At the same time,the existing mainstream depth face recognition network framework is improved,and effective model training can be realized on the small-scale sample set.This article mainly did the following work:(1)In view of the fact that the existing relatively common face detection algorithms have low recognition accuracy,this paper uses convolutional neural network for face detection,which can better detect face images and multiple face images with large deflection angles.Next,three common feature point alignment models are briefly described and feature point detection is performed using the CLM method.In this paper,we use Supervised Descent Method(SDM)algorithm to solve the problem of optimal descent of gradient descent in the process of feature point detection,and realize the fast and accurate detection of feature points.(2)This paper analyzes the current mainstream face recognition method based on deep learning,and proposes an improved volume and neural network training on the smaller-scale data set constructed in this paper,which is used to correct feature extraction of face.Face detection and face recognition tests were performed on LFW benchmark datasets and self-built datasets using different distance metrics to determine the best network model,distance metrics,and matching criteria.The algorithm achieves the following functions through simulation experiments: calling the camera to detect the face in real time,aligning the face after detection of feature points,and using the trained model for feature extraction and face recognition.If it is a stranger,it can register its identity.To the database.In summary,this paper adopts face detection algorithm based on deep learning to improve the stability and real-time performance of the algorithm.